Distributed hierarchical optimal routing using aggregation/disaggregation and decomposition/composition techniques

The authors improve the performance of the original iterative aggregation disaggregation gradient projection (IAD-GP) algorithm based on the idea of A/D only. The A/D concept naturally fits the hierarchical structure of the large wide area data networks and one would expect speed-up effects. Furthermore, by adding to the modified IAD-GP iteration decomposition/composition (D/C) steps, the authors propose a new algorithm called HAD (hierarchical A/D and D/C)-GP algorithm. The HAD-GP algorithm achieves significant improvement in convergence speed. This speed-up is demonstrated by numerical implementations comparing the HAD-GP algorithm with the modified IAD-GP and the ordinary GP (ORD-GP) algorithms